ISO/IEC TR 15938-8:2002/Amd 5:2010

Information technology — Multimedia content description interface — Part 8: Extraction and use of MPEG-7 descriptions — Amendment 5: Extraction and matching of image signature tools

Technologies de l'information — Interface de description du contenu multimédia — Partie 8: Extraction et utilisation des descriptions MPEG-7 — Amendement 5: Extraction et correspondance des outils de signature d'image

General Information

• Status: Published
• Publication Date: 10-Mar-2010
• ICS: 35.040 - Information coding
• ICS: 35.040.40 - Coding of audio, video, multimedia and hypermedia information
• Technical Committee: ISO/IEC JTC 1/SC 29 - Coding of audio, picture, multimedia and hypermedia information
• Drafting Committee: ISO/IEC JTC 1/SC 29 - Coding of audio, picture, multimedia and hypermedia information
• Current Stage: 6060 - International Standard published
• Start Date: 11-Mar-2010
• Due Date: 13-Jan-2012
• Completion Date: 13-Jan-2012

Relations

• Amends: ISO/IEC TR 15938-8:2002 - Information technology — Multimedia content description interface — Part 8: Extraction and use of MPEG-7 descriptions
• Effective Date: 25-Dec-2021

Overview

ISO/IEC TR 15938-8:2002/Amd 5:2010 is an important amendment to the ISO/IEC 15938-8 standard, focusing on the extraction and matching of image signature tools. It extends the multimedia content description interface, specifically targeting MPEG-7 descriptions for image signature extraction and similarity matching. This amendment introduces robust, compact, and efficient visual signatures designed for identifying duplicate and modified images with high accuracy.

The image signature approach provides a fingerprint-like mechanism, enabling unique identification of individual images across various common image editing operations. It incorporates global and local signatures to support comprehensive and scalable image matching.

Key Topics

• Image Signature Components
  The image signature is composed of three key parts:

  • Two global signatures representing the entire image.
  • A set of local signatures capturing localized image features.

• Feature Extraction
  Extraction techniques are based on methodologies outlined in ISO/IEC 15938-3:2002/Amd.3:2009, enabling reproducible generation of unique image signatures.

• Similarity Matching Techniques

  • Global Signatures: Matching is conducted through normalized Hamming distance on binary bit vectors representing the signatures. Threshold values are provided to minimize false positive matches dramatically.
  • Local Signatures: Matching involves a four-stage process combining hypothesis formation with geometric constraints such as feature direction, line length ratios, and area ratios for robust verification.

• Hashing for Performance
  The amendment proposes the use of 32-bit hashes derived from 512-bit global image signatures, and 8-bit hashes for local signatures to accelerate search and comparison processes without compromising accuracy.

• Geometric Constraints
  Robust geometric tests-including angle differences between features and line length ratios-ensure that matches are invariant to rotation, scale, reflection, and affine transformations, thereby improving false acceptance rates.

• Thresholds and Accuracy
  Suggested threshold values and procedures reduce false alarms to less than 0.05 parts per million, enabling highly reliable image identification suitable for critical multimedia applications.

Applications

• Image Duplicate Detection
  Quickly find exact and near-duplicate images in large multimedia databases regardless of common edits such as scaling or rotation.

• Copyright Protection
  Reliable fingerprinting of original images to prevent unauthorized duplication and modification.

• Digital Asset Management
  Efficient cataloging and retrieval of multimedia content using unique image signatures.

• Content-Based Image Retrieval (CBIR)
  Supports search engines and multimedia systems in matching images based on content rather than metadata.

• Forensic Analysis
  Verification and matching of images in investigative scenarios requiring high confidence in source identification.

Related Standards

• ISO/IEC 15938-3
  Details core multimedia content description tools, including feature extraction processes utilized in Amendment 5.

• ISO/IEC 15938 (MPEG-7)
  Provides a framework for multimedia content description, which Amendment 5 extends with enhanced image signature techniques.

• ISO/IEC JTC 1/SC 29
  The technical committee responsible for developments in coding of audio, picture, multimedia, and hypermedia information, under which this standard amendment was developed.

---

Keywords: MPEG-7 image signature, ISO/IEC 15938-8 amendment, image duplicate detection, multimedia content description, content-based image retrieval, image matching, global and local image signatures, feature extraction, geometric constraints, Hamming distance, image hashing, digital asset management, copyright protection, forensic image analysis.

Standards Content (Sample)

TECHNICAL ISO/IEC
REPORT TR
15938-8
First edition
2002-12-15
AMENDMENT 5
2010-03-15
Information technology — Multimedia
content description interface —
Part 8:
Extraction and use of MPEG-7
descriptions
AMENDMENT 5: Extraction and matching of
image signature tools
Technologies de l'information — Interface de description du contenu
multimédia —
Partie 8: Extraction et utilisation des descriptions MPEG-7
AMENDEMENT 5: Extraction et correspondance des outils de signature
d'image
Reference number
ISO/IEC TR 15938-8:2002/Amd.5:2010(E)
©
ISO/IEC 2010
ISO/IEC TR 15938-8:2002/Amd.5:2010(E)
PDF disclaimer
This PDF file may contain embedded typefaces. In accordance with Adobe's licensing policy, this file may be printed or viewed but
shall not be edited unless the typefaces which are embedded are licensed to and installed on the computer performing the editing. In
downloading this file, parties accept therein the responsibility of not infringing Adobe's licensing policy. The ISO Central Secretariat
accepts no liability in this area.
Adobe is a trademark of Adobe Systems Incorporated.
Details of the software products used to create this PDF file can be found in the General Info relative to the file; the PDF-creation
parameters were optimized for printing. Every care has been taken to ensure that the file is suitable for use by ISO member bodies. In
the unlikely event that a problem relating to it is found, please inform the Central Secretariat at the address given below.

©  ISO/IEC 2010
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or
ISO's member body in the country of the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail [e-mail]
Web www.iso.org
Published in Switzerland
ii © ISO/IEC 2010 – All rights reserved

ISO/IEC TR 15938-8:2002/Amd.5:2010(E)
Foreword
ISO (the International Organization for Standardization) and IEC (the International Electrotechnical
Commission) form the specialized system for worldwide standardization. National bodies that are members of
ISO or IEC participate in the development of International Standards through technical committees
established by the respective organization to deal with particular fields of technical activity. ISO and IEC
technical committees collaborate in fields of mutual interest. Other international organizations, governmental
and non-governmental, in liaison with ISO and IEC, also take part in the work. In the field of information
technology, ISO and IEC have established a joint technical committee, ISO/IEC JTC 1.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of the joint technical committee is to prepare International Standards. Draft International
Standards adopted by the joint technical committee are circulated to national bodies for voting. Publication as
an International Standard requires approval by at least 75 % of the national bodies casting a vote.
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO and IEC shall not be held responsible for identifying any or all such patent rights.
Amendment 5 to ISO/IEC TR 15938-8:2002 was prepared by Technical Committee ISO/IEC JTC 1,
Information technology, Subcommittee SC 29, Coding of audio, picture, multimedia and hypermedia
information.
© ISO/IEC 2010 – All rights reserved iii

ISO/IEC TR 15938-8:2002/Amd.5:2010(E)

Information technology — Multimedia content description
interface —
Part 8:
Extraction and use of MPEG-7 descriptions
AMENDMENT 5: Extraction and matching of image signature tools
Add after 4.8:
4.9 Visual Signatures
4.9.1 Image Signature
The visual content descriptors in ISO/IEC 15938-3 clauses 6 to 9 are very useful when trying to find images
with similar content. However, such descriptors are intended to be general and are found to be unsuitable for
the task of finding duplicate images. The image signature descriptor is robust (unchanging) across a wide
range of common editing operations, but is sufficiently different for every item of "original" content to identify it
uniquely and reliably – just like human fingerprints.
There are three components within the image signature, the first two are global signatures, representing the
complete image and the third is composed of a set of local signatures, each representing part of the image.
4.9.1.1 Feature Extraction
11.3.5 and 11.3.6 of ISO/IEC 15938-3:2002/Amd.3:2009 describe the extraction of the image signature.
4.9.1.2 Similarity Matching
4.9.1.2.1 Global Signatures
To perform matching between two global signatures B and B both of length N, the Hamming distance
1 2
should be taken:
H(,BB)=⊗B B ,
12 ∑ 1 2
where ⊗ denotes the exclusive OR (XOR) operator.
This can be normalised to the range 0-1 using the normalised Hamming distance:
ˆ
H(,BB)=⊗B B .
12 ∑ 1 2
N
Evaluation on test data has shown that for independent images a Hamming distance H(,BB) of 147
corresponds to a false alarm rate of less than 0.05 parts per million (ppm). A distance of 169 corresponds to a
false alarm rate of 1 ppm.
© ISO/IEC 2010 – All rights reserved 1

ISO/IEC TR 15938-8:2002/Amd.5:2010(E)
4.9.1.2.2 Local Signatures
1. Form hypotheses of 3 features
Use 32 out of 80 features
2. Apply Geometric Constraints
Feature direction & ratio of line length
3. Form hypotheses of 4 features
Use all 80 features
4. Apply Geometric Constraint
Ratio of areas
Figure AMD5.1 — Feature matching is carried out in a four stage process which combines hypothesis
forming (using the local signature) and geometric constraints
Matching images can be performed by comparing the local signatures which form a part of the image
signature. For efficiency four stages can be used, as shown in Figure AMD5.1, to match local signatures from
a pair of images. Hypotheses can be formed in stages one and three. A series of geometric test are performed
in stages two and four, these tests should be passed in order for a hypothesis to progress to the next stage.
The stages become increasingly computationally complex so each stage aims to minimise the number of
hypotheses that are accepted for subsequent processing.
The first stage involves forming hypotheses for potential matching feature points by comparing the first 32
feature points from a first image with the first 32 feature points from a second image. To perform matching
between feature points the local signatures (B and B ) for the feature points can be compared using the
1 2
Hamming distance:
H (B ,B ) = B ⊗B .
1 2 ∑ 1 2
A list of feature point pairs with a Hamming distance less than a predefined threshold T is created. A set of
A
hypotheses (candidate pairs of potentially matching features) is generated by taking all combinations of three
pairs of matching feature points from the list. A hypothesis is made up of a set of three pairs of feature points
( (a ,b ),(ab, ),(ab, )), where aa,,a are feature points from a first image and bb,,b are feature
11 2 2 3 3 12 3 12 3
points from a second image. A set of three pairs of feature points is declared a match if the cumulative
Hamming distance between the local signatures, of corresponding feature points, is below a constant
threshold T . The set of hypotheses is ordered by their cumulative distance, with the lowest distance (i.e.
B
highest likelihood of being a match) first. In order to minimise complexity the number of hypotheses may be
limited to the hypotheses corresponding to the lowest distance.
A second stage applies geometric constraints to each hypothesis generated from stage one.
...